Micron starts sending first 36GB HBM4 modules to his partners

Micron announced this week that it is already sending the first samples of its 36GB HBM4 modules to some of its main customers. Like the current HBM3E generation of technology, they are components focused mainly on AI accelerators.

Micron’s HBM4 memories are done by the Dram process that the company calls 1ß (1-Beta), with advanced 12-layer packaging technology. The result is a 60% increase in performance, according to the company, compared to the previous generation.

The first memory components being sent to customers use 2.048-bit interface, with total speeds exceeding 2TB/s by memory stack. In addition, Micron also promises 20% more energy efficiency compared to HBM3E.

Micron says HBM4 is “crucial” for general

High Bandwith Memory modules have been under development for years, but the advent of the Generative AI has been a great driver of its popularization, because technology will enjoy its advantages well. While companies compete to offer the most advanced versions of HBM, Micron calls “crucial” the arrival of its new HBM4 components for AI accelerators.

“Performance, the highest band and leading energy efficiency in the Micron HBM4 industry are proof of our leadership in memory technology and product leadership,” said Raj Narasimhan, one of Micon’s executives. “Relying on the memorable achievements that came with our HBM3E delivery, we continue to boost innovation with HBM4 and our robust memory portfolio and storage solutions.”

Nvidia and AMD are among Micron’s first customers

Micron does not openly talk about what their first partners are receiving samples from its 12 -layer and 36GB HBM4 modules, but some of the companies already know for other disclosures.

Read more:

Videocardz staff highlights, for example, AMD and Nvidia, who have already ensured technology orders for next generations of their GPUS for AI processing. We have the Vera Rubin architecture in the case of Nvidia and the Instinct Mi400 generation for AMD.